Radiomic features of prostate bi-parametric magnetic resonance imaging (BPMRI) associate with decipher score

Abstract

Embodiments facilitate predicting a patient prostate cancer (PCa) DECIPHER risk group. A first set of embodiments relates to training of a machine learning classifier to compute a probability that a patient is a member of a DECIPHER low/intermediate risk group based on radiomic features extracted from bi-parametric magnetic resonance imaging (bpMRI) images. A second set of embodiments relates to classifying a patient as a member of DECIPHER low/intermediate risk group, or DECIPHER high-risk group, based on radiomic features extracted from bpMRI imagery of the patient.

What is claimed is: 
     
       1. A non-transitory computer-readable storage device storing computer-executable instructions that when executed cause a processor to perform operations, the operations comprising:
 accessing a radiological image of a region of interest (ROI) demonstrating prostate cancer (PCa), where the ROI includes a tumoral region, where the image is associated with a patient; 
 segmenting the tumoral region represented in the image; 
 extracting a set of radiomic features from the segmented tumoral region; 
 providing the set of radiomic features to a machine learning classifier trained to predict DECIPHER risk group based on the set of radiomic features; 
 receiving, from the machine learning classifier, a probability that the patient is a member of a first DECIPHER risk group; 
 classifying the patient as a member of a first DECIPHER risk group or a second, different DECIPHER risk group based, at least in part, on the probability; and 
 displaying the classification, 
 where the radiological image is a bi-parametric magnetic resonance imaging (bpMRI) image, the bpMRI image including a T2W MRI image and an apparent diffusion coefficient (ADC) map of the ROI, and 
 where the set of radiomic features includes at least one ADC co-occurrence of local anisotropic gradient (CoLIAGe) feature, at least one ADC Laws features, at least one ADC Gabor feature, and at least one T2WI CoLIAGe feature. 
 
     
     
       2. The non-transitory computer-readable storage device of  claim 1 , where the set of radiomic features includes fifteen radiomic features. 
     
     
       3. The non-transitory computer-readable storage device of  claim 1 , where the machine learning classifier is a logistic regression model classifier. 
     
     
       4. The non-transitory computer-readable storage device of  claim 1 , where the first DECIPHER risk group is a DECIPHER low-risk group or a DECIPHER low/intermediate-risk group, and where the second, different DECIPHER risk group is a DECIPHER high-risk group. 
     
     
       5. The non-transitory computer-readable storage device of  claim 1 , the operations further comprising training the machine learning classifier. 
     
     
       6. The non-transitory computer-readable storage device of  claim 1 , the operations further comprising:
 generating a personalized treatment plan based, at least in part, on the classification; and 
 displaying the personalized treatment plan. 
 
     
     
       7. The non-transitory computer-readable storage device of  claim 5 , where training the machine learning classifier comprises:
 accessing a first dataset, where the first dataset includes a plurality of pre-operative bi-parametric magnetic resonance imaging (bpMRI) images of tissue demonstrating PCa in patients who underwent radical prostatectomy (RP) followed by DECIPHER tests, where a bpMRI image includes a T2W MRI image of a region of tissue demonstrating PCa, and an ADC map of the region of tissue, where the first dataset further includes, for each bpMRI image, a hematoxylin and eosin (H&E) stained image of the region of tissue represented in the bpMRI image, where the DECIPHER risk group of each patient is known, where the first dataset includes a low/intermediate DECIPHER risk group, and a high DECIPHER risk group; 
 co-registering the bpMRI imagery with the corresponding H&E imagery; 
 extracting a set of radiomic features from the first dataset, where, for each member of the plurality of pre-operative bpMRI images, the set of radiomic features includes at least one radiomic feature extracted from a T2WI image, and at least one radiomic feature extracted from an ADC map; 
 dividing the first dataset into a training set and disjoint, testing set, where the training set and the testing set include equal numbers of low/intermediate DECIPHER risk group patients and high DECIPHER risk group patients respectively; and 
 training the machine learning classifier using the training set. 
 
     
     
       8. The non-transitory computer-readable storage device of  claim 7 , where training the machine learning classifier comprises training the machine learning classifier with elastic-net regularization via a 5-fold cross validation approach. 
     
     
       9. The non-transitory computer-readable storage device of  claim 7 , where extracting the set of radiomic features includes selecting the N most discriminative radiomic features, N being a positive integer. 
     
     
       10. The non-transitory computer-readable storage device of  claim 7 , the operations further comprising testing the machine learning classifier on the testing set. 
     
     
       11. The non-transitory computer-readable storage device of  claim 9 , where the N most discriminative radiomic features are selected using a Pearson's correlation coefficient feature selection approach. 
     
     
       12. The non-transitory computer-readable storage device of  claim 9 , where the N most discriminative radiomic features are selected simultaneously with training the machine learning classifier. 
     
     
       13. An apparatus comprising:
 a processor; 
 a memory configured to store a bi-parametric magnetic resonance imaging (bpMRI) image associated with a patient, where the image includes a region of interest (ROI) demonstrating prostate cancer (PCa) pathology, the bpMRI image having a plurality of pixels, a pixel having an intensity, the bpMRI image comprising a T2 W MRI image and an apparent diffusion coefficient (ADC) map; 
 an input/output (I/O) interface; 
 a set of circuits; and 
 an interface that connects the processor, the memory, the I/O interface, and the set of circuits, the set of circuits comprising: 
 an image acquisition circuit configured to:
 access the bpMRI image; 
 
 a tumor segmentation circuit configured to:
 segment a tumoral region represented in the bpMRI image, where segmenting the tumoral region includes defining a tumoral boundary; 
 
 a radiomic feature circuit configured to:
 extract a set of radiomic features from the tumoral region represented in the bpMRI image, where the set of radiomic features includes at least one ADC co-occurrence of local anisotropic gradient (CoLIAGe) feature, at least one ADC Laws features, at least one ADC Gabor feature, and at least one T2WI CoLIAGe feature; 
 
 a DECIPHER risk group prediction circuit configured to:
 compute a probability that the patient associated with the image is a member of a first DECIPHER risk group, or a member of a second, different DECIPHER risk group, based on the set of radiomic features; 
 generate a classification of the patient as a member of the first DECIPHER risk group, or a member of the second, different DECIPHER risk group based, at least in part, on the probability; and 
 
 a display circuit configured to display the classification. 
 
     
     
       14. The apparatus of  claim 13 , where the set of radiomic features includes fifteen radiomic features. 
     
     
       15. The apparatus of  claim 13 , where the DECIPHER risk group prediction circuit is configured to compute the probability or generate the classification using a logistic regression model machine learning approach. 
     
     
       16. The apparatus of  claim 13 , where the set of circuits further comprises:
 a PCa personalized treatment plan circuit configured to:
 generate a personalized treatment plan based, at least in part, on the classification; and 
 
 where the display circuit is further configured to display the personalized treatment plan. 
 
     
     
       17. The apparatus of  claim 13 , where the set of circuits further comprises:
 a training and testing circuit configured to:
 train the DECIPHER risk group prediction circuit on a training cohort; and optionally 
 test the DECIPHER risk group prediction circuit on a testing cohort. 
 
 
     
     
       18. A non-transitory computer-readable storage device storing computer-executable instructions that when executed cause a processor to perform operations, the operations comprising:
 accessing a bi-parametric magnetic resonance imaging (bpMRI) image of a region of interest (ROI) demonstrating prostate cancer (PCa), where the ROI includes a tumoral region, where the bpMRI image is associated with a patient, the bpMRI image including a T2 W MRI image and an apparent diffusion coefficient (ADC) map of the ROI; 
 segmenting the tumoral region represented in the bpMRI image; 
 extracting a set of fifteen radiomic features from the segmented tumoral region, where the set of fifteen radiomic features includes at least one ADC co-occurrence of local anisotropic gradient (CoLIAGe) feature, at least one ADC Laws features, at least one ADC Gabor feature, and at least one T2WI CoLIAGe feature; 
 providing the set of radiomic features to a logistic regression model machine learning classifier trained to predict DECIPHER risk group based on the set of radiomic features; 
 receiving, from the machine learning classifier, a probability that the patient is a member of a first DECIPHER risk group; 
 classifying the patient as a member of the first DECIPHER risk group or a second, different DECIPHER risk group based, at least in part, on the probability, where the first DECIPHER risk group is a DECIPHER low/intermediate risk group, and where the second DECIPHER risk group a DECIPHER high-risk group; and 
 displaying the classification and optionally displaying the probability, the set of radiomic features, or the bpMRI image.